📢‼️Just published on ArXiv the first complete draft of "PoSSUM: A Protocol for Surveying Social-media Users with Multimodal LLMs” 🤖 https://t.co/Pys4p8vqA4 👇👇 1/n

3) I have so much unpublished stuff it is scary. https://t.co/D4vK4nubRM

AI always calling your ideas “fantastic” can feel inauthentic, but what are sycophancy’s deeper harms? We find that in the common use case of seeking AI advice on interpersonal situations—specifically conflicts—sycophancy makes people feel more right & less willing to apologize.

@gianlubaio this could be of interest !

Sophie first presented this at the @mcmc_stan conference in Oxford, where we received valuable feedback from the brilliant Stan community. Without Stan, we could not have done this research.

4/4 We also quantify the nanoparticulate fraction of E171 🧫 Pre-ban: median exposure was 0.0274 mg/kg/day (95th 0.0819) ‼️We find it is far above earlier estimates that assumed a tiny nano share‼️ Market presence matters too across foods, meds, toothpaste & lip balm 📈🛍️

3/4 TiO₂ in Belgium 🇧🇪 Pre-ban: median exposure was 0.1139 mg/kg/day (95th 0.3624) mostly from food (87%) highest for Kids Post-ban (counterfactual): median 0.0108 (95th 0.0483) mostly from medicine (81%) highest for the Elderly (65+) 👶🍭➡️👵💊

2/4 We present a novel Hierarchical Bayesian framework for estimating chemical exposure 🧪📊in a population. We merge food consumptions surveys, lab concentrations, market presence & demographics and simulate to generate representative exposure estimates

📢Preprint Alert‼️ https://t.co/Bg9eLxZW5u Who is most exposed to banned chemical Titanium Dioxide ? Before controversial EU ban, it was kids under the age of 10 !🍬 Excited to see this important method + substantive contribution by my student Sophie (@sciensano + @data_um)👇1/4

By using large language models as R data.table coding assistants, there are even fewer reasons to avoid the elegance and speed of data.table for data manipulation/aggregation/complex longitudinal data processing: https://t.co/YlcWJsHGyH #RStats

Stan v2.37. Thanks to everyone who contributed! https://t.co/SXbGslTvkP

Excited to share our new paper in Political Analysis co-authored with Neil Ketchley, @aasiegel, and Mossaab Bagdouri! 📢 We introduce an open-source, computationally light method for measuring media criticism—a key marker of media freedom and democratization.

Open Research (OpenAI-linked research outfit) just put out a new report on using LLMs for public opinion polling that suggests they may improve response rates substantially:

Folks, I ran the numbers on the UK government's recommendation to delete old photos and emails to save water. Link below

Akshually Italy is an export-oriented, high-tech manufacturing economy that even by economists' standards is larger than the UK's despite a smaller population. Italy is a major machine tool exporter. Tesla got its giga-presses from a boutique Italian firm outside of Milan.

https://t.co/rKRSvYIcv6

World War 1 Austrian propaganda poster portraying the Italians as sea monsters attempting to occupy Dalmatia.

Finishing my slides for a talk on the Elements of Differentiable Programming this Wednesday at @lossfunk Most material is from the book by @mblondel_ml and @vincentrouletx at @GoogleDeepMind. You can attend in-person or online. Link 👇

Happy Father's Day https://t.co/Mwx25RMQQQ

In our competitive society, where it seems that only the strong and winners deserve to live, sport also teaches us how to lose. It forces us, in learning the art of losing, to confront our fragility, our limitations and our imperfections. It is through the experience of these

this absolute maniac implemented the entire diffusers library… from scratch… in R didnt know that was possible (does it run on GPU?). and i dont know if i love it or hate it. but god do i respect it